Is AI a Bubble Right Now? The DN AI Bubble Gauge for 2026
AI Bubble or Real Revolution? The 5 Signals Every Investor Should Watch
Is AI a Bubble Right Now? The DN AI Bubble Gauge for 2026
Five measurable signals of froth, distilled into one score — the scoreboard the most-argued question in markets has been missing.
Whether AI is a bubble is not a yes-or-no question but a matter of degree, measurable across five signals: how far AI capital spending outruns the revenue it generates, the gap between that spending and realised returns, how stretched AI valuations are versus history, how concentrated the market has become in a handful of AI names, and how frantic private AI funding is. The DN AI Bubble Score below weights these into a single 0–100 reading, where under 25 means froth is low and over 85 means peak mania. It is a transparent framework you set to current conditions — not a verdict that AI is or isn't a bubble, but a way to measure how much one is forming.
No question in markets is argued more often, or settled less, than whether artificial intelligence is a bubble. One camp points to a genuine technological revolution with real, fast-growing revenue and insists this time is different. The other points to hundreds of billions in capital spending chasing a fraction of that in actual AI income, to indices dangerously concentrated in a few names, and to the unmistakable rhyme of every bubble before it. Both sides marshal selective evidence, and the debate goes in circles because there has never been an agreed scoreboard.
This gauge is that scoreboard. It does not declare a winner — declaring AI definitively a bubble or definitively not is exactly the overconfidence that gets people hurt at both tops and bottoms. Instead it breaks the question into five measurable components, lets you set each to current conditions, and returns a single DN AI Bubble Score that says, in one number, how much froth the evidence actually supports. When the debate next reignites, this is the reference that grounds it.
The anatomy of a bubble
Bubbles are not simply high prices; they are a specific condition in which price detaches from the cash flows that could ever justify it, sustained by the belief that someone will pay more tomorrow. Every great bubble shares a skeleton: a genuine innovation that warrants real excitement, a flood of capital that overshoots what the innovation can yet absorb, valuations that stop referencing fundamentals, a narrowing of the market into a few darlings, and a final phase where funding and belief feed on each other. The danger is that the genuine innovation at the centre makes the excess feel justified right up until it doesn't.
That is what makes AI so hard to judge. The technology is unmistakably real and its revenue is genuinely growing fast, which is precisely the condition under which the most dangerous bubbles form — nobody inflates a mania around something worthless. The question is never whether AI matters; it obviously does. The question is whether the capital being deployed has run ahead of the returns that capital can plausibly earn, and by how much. The gauge measures that distance.
The five inputs
Each signal captures one symptom of froth, and each is something you can read from current reporting and set on the gauge:
How far AI capital spending — data centres, chips, power — outruns the revenue AI currently generates. Building modestly ahead of demand is normal; spending several times current revenue is the classic signature of capital racing past fundamentals.
The crux. The distance between the enormous sums being invested and the realised, durable profit those investments are actually producing. A wide, persistent gap is what eventually forces a reckoning.
How stretched AI-leader valuations are against their own history and the broader market. Rich multiples can be earned by growth, but the further they run, the less margin for disappointment remains.
How much of the major indices now rests on a handful of AI-exposed mega-caps. Extreme concentration means the whole market's fate is tied to a few stories staying intact.
How frantic private AI funding has become — round sizes, valuations and speed. Reflexive, accelerating funding that prices startups on narrative rather than traction is late-cycle behaviour.
Reading the score
The composite runs from zero to one hundred, rising with froth. It is deliberately not a buy or sell signal — a high reading can persist for a long time, and bubbles inflate far longer than sceptics expect, while a low reading does not mean prices cannot fall for other reasons. What the score offers is calibration: a way to hold a view about AI risk that is proportional to the actual evidence rather than to the last headline you read.
Froth is low; prices broadly track fundamentals and capital is disciplined.
Enthusiasm is building and excess is forming, though still partly justified by real growth.
Classic bubble signatures appear across several signals at once. Late-cycle behaviour dominates.
Extreme froth across the board — the kind of reading that has historically preceded sharp corrections.
The gauge also names the hottest and coolest signals, which matters more than the headline number. A high score driven mostly by valuation is a different animal from one driven by a yawning spend-to-return gap; the former can deflate gently through time and growth, while the latter tends to break suddenly when an earnings season fails to deliver. Knowing which signal is loudest tells you what to watch.
The bull and bear cases, fairly stated
Authority demands holding both sides honestly, because both are serious:
Why it may not be a bubble
AI revenue is real and compounding at rates few technologies have matched. The spenders are largely profitable incumbents funding capex from cash flow, not speculative debt. Productivity gains are showing up in real workflows, and prior technology build-outs that looked like overspending — railways, fibre — ultimately underpinned decades of growth even as early investors were wiped out.
Why it may be
Capital expenditure dwarfs realised AI revenue, and much of the spend is circular, with the same few players funding each other. Market concentration is at historic extremes, valuations price near-flawless execution, and the productivity case remains more promised than proven at scale. Every bubble felt justified by a real revolution at the time.
What bursts it
Bubbles do not end because they become expensive; they end when the marginal buyer runs out or the story breaks. For AI, the most-watched pin is an earnings cycle in which the vast capital spending visibly fails to convert into commensurate profit, forcing the market to reprice the spend-to-return gap all at once. Other candidates include a funding freeze that strands cash-burning startups, a stumble in one of the mega-caps that cascades through over-concentrated indices, or a macro shock — a rate move or growth scare — that re-rates stretched multiples. The gauge cannot time the pin, but by surfacing which signal is hottest, it tells you where to keep your eyes.
Where to act on a view
This is a framework for calibrating risk, not a trade signal — but if you hold a view, you need the tools to express it. To screen AI-exposed equities and track the comparables and concentration yourself, charting platforms are the natural home; for the crypto-native slice of the AI narrative, AI tokens trade on the major exchanges.
Frequently asked questions
Is AI a bubble right now?
It is best treated as a matter of degree rather than yes or no. The DN AI Bubble Gauge measures it across five signals — capex versus revenue, the spend-to-return gap, valuation richness, market concentration and funding velocity — and returns a 0–100 score. Set each input to current conditions to see how much froth the evidence supports; under 25 is low, over 85 is extreme.
What is the DN AI Bubble Score?
It is a composite 0–100 score from Decentralised News that weights five measurable froth signals: capex-to-revenue and the spend-to-return gap at 25% each, valuation richness at 20%, and market concentration and funding velocity at 15% each. Higher means more bubble-like conditions.
How is the AI boom different from the dot-com bubble?
The parallels are real — concentration, capital intensity, narrative-driven funding — but so are the differences. Much of today's AI spending comes from highly profitable incumbents funding capex from cash flow rather than speculative debt, and AI revenue is growing faster than internet revenue did in 1999. That can make froth more durable, but does not eliminate the risk that spending has outrun returns.
What would burst an AI bubble?
Most likely an earnings cycle where massive AI capital spending visibly fails to produce commensurate profit, forcing a sudden repricing of the spend-to-return gap. Other triggers include a funding freeze stranding cash-burning startups, a stumble in a dominant mega-cap cascading through concentrated indices, or a macro shock that re-rates stretched valuations.
Does a high bubble score mean I should sell?
No. It is a calibration tool, not a trade signal. Bubbles can inflate far longer than sceptics expect, and a high reading can persist for a long time. The score helps you size risk in proportion to evidence; timing any action still requires your own judgement and risk management.
Is AI capex really too high?
That is the central debate, captured in the capex-to-revenue and spend-to-return signals. Bulls argue the build-out underpins decades of future growth, as railways and fibre eventually did; bears note that capital expenditure currently dwarfs realised AI revenue and much of it is circular. The gauge lets you weigh the evidence rather than pick a side by instinct.
This tool and article are for educational and informational purposes only and do not constitute financial, investment or trading advice, nor a prediction about AI valuations or any security. The DN AI Bubble Score is a transparent model based on subjective, user-set inputs and is not a forecast; froth can persist or unwind unpredictably. Markets carry significant risk including loss of capital. Always do your own research and consider consulting a licensed financial professional. Decentralised News may earn a commission from services linked in this article at no additional cost to you.
